**** BEGIN LOGGING AT Mon Mar 18 03:00:21 2019 Mar 18 07:46:23 good morning Mar 18 08:41:36 Hi everyone Mar 18 08:42:14 Is here anyone that could help my with a simple question? Mar 18 08:44:09 I have build an image with yocto, then after building it I tried to add other packages both using bitbake command and also editing locl.conf file but after building the image another time nothings seems to have changed, in other words the packages are not added to the new image Mar 18 08:44:42 Is there a way to add packages after having compiled an image or should I recompile everything every time? Mar 18 08:46:30 SimoneNascivera: you have to rebuild the image, yes Mar 18 08:46:34 SimoneNascivera: first of all ensure you added the packege using IMAGE_INSTALL_append = " packagename" Mar 18 08:47:11 SimoneNascivera: you have to add whatever you need to the image, be it through IMAGE_INSTALL, CORE_IMAGE_EXTRA_INSTALL (or howsit called), and then rebuild the image. Mar 18 08:47:54 SimoneNascivera: then you can call: bitbake yourimagename -C rootfs that Populate rootfs again Mar 18 08:48:07 or just bitbake your_image Mar 18 08:49:47 SimoneNascivera: you have to be aware that this won't recompile everything, it reuses as much as possible. if all packages are already built, this actually just "repackages" everything into the image Mar 18 08:53:20 mckoan: bitbake yourimagename -C rootfs seems to work Mar 18 08:53:38 mckoan: thank you so much Mar 18 08:59:07 mckoan: the command did actually rebuild the rootfs and image file, but it did not add the package to it, even if I added it both in local.conf and using bitbake Mar 18 09:11:01 New news from stackoverflow: create the symlink failed in yocto project Mar 18 09:27:12 armpit: there is a build that consists of bitbake and oe-core, yes. Sorry its causing so much pain :/ Mar 18 09:51:31 Hi guys, what is the equivalent of "bitbake virtual/kernel -c menuconfig" for devtool ? Mar 18 09:51:38 is there one ? Mar 18 09:52:04 In the end I just want to access linux menuconfig from within devtool Mar 18 10:41:21 yacar_: it is a not available option, there is a patch in waiting list: https://patchwork.openembedded.org/patch/141208/ Mar 18 10:42:02 Ok thanks for the info, but I guess modifying directly my kernel's .config should also do the trick ? Mar 18 10:42:40 yacar_: what's the problem using bitbake ? Mar 18 10:44:33 It's not really a problem here, it's just for convenience. My use case is this: up to now I was only modifying stuff in the kernel using the SDK, but now that everything is set up I want to add a new driver directly in the kernel and start the development. Mar 18 10:46:58 Plus I'm not exactly familiar with the gymnastic of going back and forth from yocto to the SDK. Mar 18 10:47:35 Dear All, Mar 18 10:47:35 I would like to ask for a confirmation if my understanding is correct Mar 18 10:48:16 The libstd++ is provided in OE/Yocto as part of gcc-runtime recipe (provided with gcc recipe to be used on target) Mar 18 10:48:43 Hence the only way to update it (similar to one here: https://forum.linuxgameconsortium.com/t/libstdc-so-6-version-glibcxx-3-4-22-not-found-fix/316) Mar 18 10:49:03 is to update the recipe, which provide gcc -> e.g. from 6.x to 7.x Mar 18 10:49:05 yacar_: if you are building a kernel module, you can do that out of the kernel tree at the beginning Mar 18 10:49:26 yacar_: see my presentation "Kernel Modules with eSDK" https://youtu.be/K3Jqvv5q5eI Mar 18 10:50:00 mckoan: Thanks :) Mar 18 11:19:37 how i modify recipe added with bakekake-layers add-layer Mar 18 11:20:09 using devtool modify ? Mar 18 11:45:32 eSDK dosen't automatically update with poky 7e7ee662f5dea4d090293045f7498093322802cc Mar 18 11:51:57 lukma: yes, libstdc++ comes from gcc and you want it to match your compiler Mar 18 12:41:40 New news from stackoverflow: how to make bitbake print options of do_configure Mar 18 14:03:18 Hi community, i would like to change my kernel version on my build, i make the following change on my local.conf: PREFERRED_VERSION_linux-karo = "4.13%" Mar 18 14:03:50 But i got the following error: Multiple versions of linux-karo are due to be built (/home/theoris/projects/EdDUV2_yocto/fsl-community-bsp/sources/meta-fsl-arm-extra/recipes-kernel/linux/linux-karo_4.13.bb /home/theoris/projects/EdDUV2_yocto/fsl-community-bsp/sources/meta-fsl-arm-extra/recipes-kernel/linux/linux-karo_4.1.15.bb). Only one version of a given PN should be built in any given build. You likely need to set PREFERRED_VERSION Mar 18 14:03:59 to select the correct version or don't depend on multiple versions Mar 18 14:04:30 Any idea on how to fix this, thanks in advance ! Mar 18 14:08:21 vladzouth: you can't put PREFERRED_VERSION_linux-karo in local.conf, set it into the MACHINE config file Mar 18 14:09:57 you can generally, unless this bsp is being dumb Mar 18 14:10:15 https://github.com/Freescale/meta-freescale-3rdparty/blob/master/conf/machine/include/tx6-karo-common.inc Mar 18 14:11:23 mckoan: Ok i will try to do it in machine.conf file Mar 18 14:12:44 rburton: what you are saying is that we can do the modification in local.conf ? Mar 18 14:13:16 you *should* be able to. Mar 18 14:13:55 OK Thanks ! Mar 18 14:14:43 I try to put that in local.conf or machine.conf files but it doesn't change anything, i still have the same error Mar 18 14:14:48 quetion for the ptest-gurus around: we have a test-image that does everything nice and clean, but the result of the system ptest runner always seems to be "UNKNOWN", regardless if it runs one of our tests or an included one, and regardless if if passes or not. any pointers? Mar 18 14:15:13 LetoThe2nd: does your test output in the format ptest-runner is expecting? Mar 18 14:15:39 Hi there, I was wondering if it would be possible to enable a complete bbappend recipe for a given override (in my particular case that would be libc-musl) Mar 18 14:15:49 rburton: how to check? is it documented? Mar 18 14:16:34 rburton: its not my project, i'm just the community facing part ATM. and $coworker said that he gets "UNKOWN" also for the included test that he tried. Mar 18 14:17:32 LetoThe2nd: well, ptest-runner really doesn't care much Mar 18 14:17:42 ross@flashheart ~/Programming/ptest-runner2 (master) Mar 18 14:17:42 $ git grep UNKNOWN |wc -l Mar 18 14:17:43 0 Mar 18 14:18:07 so i think you need to find out what is printing unknown Mar 18 14:18:24 rburton: can you try for lower case or case invariant too please? Mar 18 14:18:31 then i'll forward it. Mar 18 14:18:34 $ git grep -i UNKNOWN |wc -l Mar 18 14:18:34 0 Mar 18 14:19:20 i'd love to see ptest overhauled, maybe mandate TAP-style output for the tests, so we can patch less and reuse more tooling Mar 18 14:19:27 rburton: so in a nutshell, the ptest-runner is not expected to output the literal "unknown" by itself. right? Mar 18 14:19:39 ptest-runner does not contain the string unknown Mar 18 14:19:53 maybe he means the gnome test runner? Mar 18 14:19:56 rburton: actually $coworker was explicitly pleased with the test-image, that was the only thing he wondered about :) Mar 18 14:20:34 LetoThe2nd: we've got more tooling now, RP did some great work if you're going to be running ptest images often you might want to find out what we have (because I can never remember) Mar 18 14:21:26 rburton: yeah, i was aware that there has been recent work, thats why i also asked. Mar 18 14:24:07 LetoThe2nd, rburton: UNKNOWN sounds like something coming from python unittest Mar 18 14:24:52 LetoThe2nd: ptest in testimage is handled a little differently since there is no one exit status that can tell us what we need to know. Its likely some ptest will always fail Mar 18 14:25:16 LetoThe2nd: instead we report the ptest results separately. If all the ptests pass it should show success though Mar 18 14:25:49 RP: yes exactly. $coworker says that he gets the seperate results in seperate logs, all is well. he just wondered about the "net" result. Mar 18 14:26:31 RP: any pointer i can give him to dig deeper? Mar 18 14:27:53 LetoThe2nd: what behaviour do they want? Mar 18 14:28:08 RP: "show PASS or FAIL" :) Mar 18 14:28:31 LetoThe2nd: http://git.yoctoproject.org/cgit.cgi/poky/tree/meta/lib/oeqa/runtime/cases/ptest.py Mar 18 14:28:39 * RP notes the @unittest.expectedFailure Mar 18 14:29:27 LetoThe2nd: also, https://bugzilla.yoctoproject.org/show_bug.cgi?id=13150 Mar 18 14:29:28 Bug 13150: normal, Medium+, 2.7 M3, richard.purdie, NEW , Test report contains UNKNOWN status when tests were skipped Mar 18 14:29:48 LetoThe2nd: so I suspect "expected failures" aren't being handled correctly Mar 18 14:29:52 what package provides gdbserver ? Mar 18 14:30:02 yates: gdbserver? Mar 18 14:30:07 RP: ah ok, so its basically "a bug" Mar 18 14:30:12 nm Mar 18 14:30:17 no, it's gdb Mar 18 14:30:40 yates: nm is in binutils Mar 18 14:30:54 yates: gdb is the recipe in the gdbserver package Mar 18 14:31:08 LetoThe2nd: approximately speaking Mar 18 14:32:29 RP: thanks! don't get me wrong, he explicitly pointed out that he liked the whole construction very much. its just in discussion that i told him there has been recent work, and so he asked me to check with folks here if it might be due to exactly that rework before her really goes digging deeper. Mar 18 14:34:22 LetoThe2nd: I think its somewhere between that bug I linked to and the fact we "expect" ptest to fail which has a special status in py unitest Mar 18 14:35:08 LetoThe2nd: glad they liked it and any help in improving gratefully received Mar 18 14:35:17 RP: could you elaborate on "expect to fail?" Mar 18 14:36:04 RP: now that we've got the context sorted out i'll see if we can help improve or will just live with the current state. no promises there, sorry. Mar 18 14:36:08 LetoThe2nd: https://docs.python.org/3/library/unittest.html#skipping-tests-and-expected-failures Mar 18 14:37:09 LetoThe2nd: its a unittest outcome (pass, fail, skip, expectedFail, unexpectedSuccess, unknown) Mar 18 14:38:21 RP: i see, thanks. I'll forward this, and if we can help I'll let you know. Mar 18 14:39:04 LetoThe2nd: we will likely improve things, just other priorities right now Mar 18 14:47:14 RP: whatever. i bitbake gdb, not bitbake gdbserver Mar 18 14:47:48 it generates a package gdb, which i then smart install as gdb. sounds like gdb is the package to me Mar 18 15:15:50 rburton: were you able to discern any other changes to the matchbox window manager recipe? just patching so that the --enable-composite option builds still did not provide transparency Mar 18 15:18:09 where are the latest oe recipes maintained? Mar 18 15:28:08 hello :-) Mar 18 15:28:22 I have a problem while trying to build an image with its sdk : Mar 18 15:28:32 Problem: package perl-module-overload-5.24.4-r0.armv7at2hf_neon requires perl-module-warnings-register, but none of the providers can be installed Mar 18 15:28:49 (fsl-image-machine-test from fsl-community-bsp) Mar 18 15:28:58 building the image in itself works correctly Mar 18 15:29:13 where should I look in order to be able to fix this ? Mar 18 15:38:15 I tried compiling a qemux86 image of a Qt application (non-gui). When I fire up qemu and try to run it, I get SIGILL, illegal instruction from libQt5Core.so. Any ideas in how to apprach this? Am I running my qemu correctly? Mar 18 15:39:43 I noticed that I cannot run runqemu myimage straight out of the box, its complaining about "*.qemuboot.conf: No such file or directory". When I give it the .ext4 image specifically, I get the VM prompt Mar 18 15:39:53 yates: my point is to ask about recipes, not packages then. Two different things, say what you mean! :) Mar 18 15:42:35 RP, regarding the oecore only build, it means changes have to be done in three repos Poky, core and bitbake for each and every mut and next style builds. Mar 18 15:43:28 if not, than any quick build is in accurate Mar 18 15:44:17 can we move that build to full ? Mar 18 15:44:32 runqemu fails with "runqemu - ERROR - Command 'ls -t /home/sveinse/build-qemu/build/tmp/deploy/images//*.qemuboot.conf' returned non-zero exit status 2.", which seems to me to come from "bitbake -e", so perhaps the "//" is an indication that I'm missing setting a variable in my image? Mar 18 15:47:50 armpit: I guess I've just been aware its a known weakness. My -next builds are run in three different branches Mar 18 15:48:53 does Ross's ? Mar 18 15:49:49 master-next is not part of this as its updated by you at the same time Mar 18 15:50:34 armpit: no, mut doesn't Mar 18 15:51:25 armpit: You can easily replicate my scripts/setup if it bothers you but I really would try just not to worry about that build. Just be aware of what it represents Mar 18 15:51:32 k, now that I now I will adjust my process Mar 18 15:53:44 the good thing it stagged sumo changes are good for a pull request so we can get the next point release started Mar 18 15:55:02 armpit: apart from testresult and the -helper changes? :/ Mar 18 15:56:19 err, fine... I still want the batch of my plate to then get thud sorted for those needed changes Mar 18 15:58:40 armpit: that is fine, one step at a time Mar 18 15:58:59 armpit: I'm happy to help with this, I just need M3 built using the new process/tools Mar 18 16:00:27 It is related to DEPLOY_DIR_IMAGE, which in turn is dependent on MACHINE. Now, unless MACHINE is explicitly defined (MACHINE=) in either local.conf or as enviroment, this variable is wrong. It is not correct if MACHINE is weakly defined using MACHINE?= or MACHINE??= Mar 18 16:00:47 Isn't this somewhat unexpected behaviour? Mar 18 16:12:54 has anyone ever seen the matchbox window manager compositing work so that transparency works? e.g., putting a wxPanel on a wxFrame that is transparent? (speaking in terms of wxWidgets)? Mar 18 16:13:55 rburton: have you? Mar 18 16:28:06 rburton: i see you've done some work on the latest versions of matchbox-wm - could ypu please respond to my previous question? Mar 18 16:28:24 namely, have you ever seen the matchbox window manager compositing work so that transparency works? e.g., putting a wxPanel on a wxFrame that is transparent? (speaking in terms of wxWidgets)? Mar 18 16:29:33 as i stated a couple of weeks back, this works under desktop linux (presumably because it has a decent compositor) but does NOT work with mbwm, even though i've recompiled to enable the repositor (and patched the -lm problem) Mar 18 16:30:15 yates: as i said last time, wm compositing has nothing to do with alpha blending inside the windows Mar 18 16:30:34 then what does? Mar 18 16:32:02 do you have any clue why my transparency works under desktop linux but not under yocto linux (using wxwidgets)? Mar 18 16:34:29 yates: wxwidgets Mar 18 16:34:42 because you can't get a 32-bit visual maybe? because wxwidgets doesn't have compositing enabled? Mar 18 16:35:01 what do you mean by a "32-bit visual"? Mar 18 16:35:04 *how* does wx do transparency Mar 18 16:35:19 gtk+ does it by just drawing everything onto a single X surface Mar 18 16:35:24 so it has complete control Mar 18 16:35:57 kergoth: the one thing that article is missing is a list of five blessed licenses we can all move to slowly Mar 18 16:36:39 i don't know Mar 18 16:37:21 yates: i'd be asking the wx channel what would explain your behaviour, instead of asking people who have never used wx Mar 18 16:38:17 well i have been. no one seems to know there, and i thought there may be some people here who have used wx Mar 18 16:38:43 perhaps i should just dive into the wx source Mar 18 16:39:35 and i thought you said a week or two ago yourself that the problem was because compositing was not enabled on mbwm.. Mar 18 16:39:58 you said window transparency :) Mar 18 16:40:26 oh Mar 18 16:40:32 don't even know what a wxframe *is* Mar 18 16:40:45 probably don't care, either? Mar 18 16:40:48 :) Mar 18 16:41:08 if a wxframe is a top-level then you care for window-level transparency Mar 18 16:41:25 https://docs.wxwidgets.org/3.0/classwx_frame.html Mar 18 16:41:40 if its not then you care about wx-level, but as all i know about wx is that it basically wraps other toolkits then it depends on what the toolkit can do Mar 18 16:41:41 yes, it is Mar 18 16:41:43 ie i doubt gtk2 can do that Mar 18 16:42:01 i.e., a wxFrame is the top level client area of the window Mar 18 16:43:31 perhaps it is the case that my desktop linux wx is based on gtk3 but the embedded yocto wx is based on gtk2 Mar 18 16:44:05 i see to recall this is your wx recipe Mar 18 16:44:26 gtk2 is so unmaintained and dead its not even funny, so if you're doing that then switch to gtk3 before gtk4 is released Mar 18 16:44:52 ok, right. looking into it now Mar 18 16:45:09 yes, it is my recipe Mar 18 16:45:44 did you submit it to meta-oe yet? Mar 18 16:46:15 no. i guess i should? Mar 18 16:46:21 yes Mar 18 16:46:27 RP, i just sent a revert for boost. I am starting a build just to make sure it does break something. I'll let you know if its ok Mar 18 16:46:51 armpit: "to make sure it does break something" ? :) Mar 18 16:46:59 armpit: is this a new policy? Mar 18 16:49:26 hehe Mar 18 16:49:37 :) Mar 18 16:50:08 I am finding doing things wrong unites the community Mar 18 16:50:09 i thought that was policy? Mar 18 16:50:31 unites the community against common enemy? :) Mar 18 16:50:37 ok Mar 18 16:50:45 rburton, I changed wiki so it must be true ; ) Mar 18 16:51:18 the Internet does not lie Mar 18 16:51:45 JaMa, thats right.. I am the unifier Mar 18 16:51:46 rburton: my wx recipe DEPENDS on gtk+. it's kinda confusing to me which version (2 or 3) this grabs. if these are my gtk+ recipes, which would be grabbed? https://paste.fedoraproject.org/paste/1H3iOrxeIKCJMXi25EFo5w Mar 18 16:51:51 * JaMa just noticed "Having difficulty on the list, or with someone on the list?" in the topic :) Mar 18 16:52:00 yates: thats 2 Mar 18 16:52:06 yates: the gtk3 recipe is called gtk3 Mar 18 16:52:13 not gtk+3? Mar 18 16:52:24 no, because the depends is gtk+ Mar 18 16:52:35 so you're after the recipe called gtk+_ Mar 18 16:52:42 not gtk+3_ Mar 18 16:52:57 filename is name_version Mar 18 16:54:45 armpit: I'm now wondering if each change can only break one thing or if multiple breakages are allowed if specified in the commit message :) Mar 18 16:54:46 Any quick proposals off the bat why libQt5Core SIGILLs when building it for qemux86 and running in qemu? Does it ring any bells to anyone? Mar 18 16:55:14 sveinse: wrong processor being emulated by qemu? Mar 18 16:55:46 sveinse: yes, probably wrong -cpu parameter, but it should show an warning before dying like that Mar 18 16:56:07 sveinse: because qt is ignoring the tune, trying to use cool instructions, and qemu is being told to emulate a machine that doesn't have those instructions. *might* be because qemux86 machine and qemu invocation don't match, but i thought we'd aligned those. Mar 18 16:56:11 sveinse: unless you see the SIGILL somewhere in mesa/llvm stack which is what I'm seeing now even with correct -cpu Mar 18 16:56:45 RP: yeah, I suspected. Yet, runqemu prints MACHINE: [qemux86], which I believe should be correct Mar 18 16:56:47 sveinse: something like this? https://paste.ubuntu.com/p/PDnnJGvFw8/ Mar 18 16:57:21 so i should DEPENDS gtk+_3.20.9? Mar 18 16:57:34 JaMa: alledgedly the app is a non-GUI, but I'll have to confirm that, so for now, all possible causes are open I suppose Mar 18 16:57:37 yates: no. DEPENDS += "gtk+3" Mar 18 16:58:01 yates: gtk+ and gtk+3 are two different recipe PN Mar 18 16:58:16 sveinse: FWIW by backtrace is from VirtualBox, not QEmu Mar 18 16:58:16 oh. Mar 18 17:00:46 sveinse: you can attach a gdb to the app inside qemu and watch what instruction it crashes on Mar 18 17:01:36 JaMa: no, it seems, its only dl_open and libQt5Core in the bt before dying. I need to install the -dbg packages in this image so that I can get more sensible data from it Mar 18 17:03:24 sveinse: ok, might still be the -cpu parameter, a bit more details here: http://lists.openembedded.org/pipermail/openembedded-core/2018-April/150173.html Mar 18 17:03:33 I need to debug more on this on my own. How can I know that -cpu parameter I should have for qemux86? Is there somewhere I can look at for .... right, thanks :D Mar 18 17:04:23 sveinse: are you using runqemu script? IIRC it shows the whole cmd line used to execute qemu (which includes the -cpu parameter) Mar 18 17:07:12 JaMa, yes I do, https://bpaste.net/show/91efc35c360c . Mar 18 17:08:47 -cpu pentium2 shown on last line Mar 18 17:10:45 The qt configure log is lost in the sstate cache, but AFAICR I think it does compile with sse options. I think I've seen that. Mar 18 17:10:48 which is the default for 32bit qemux86 Mar 18 17:10:49 meta/conf/machine/include/qemuboot-x86.inc:QB_CPU_x86 = "-cpu pentium2" Mar 18 17:10:49 meta/conf/machine/include/qemuboot-x86.inc:QB_CPU_KVM_x86 = "-cpu pentium2" Mar 18 17:15:13 Allright, I need to test more tomorrow. Thanks JaMa Mar 18 17:17:27 Oh, btw, I picked qemux86 just by random (and because my HW system is 32-bit arm). Is it likely that this would be fixed if I compiled the whole thing for qemux64? Mar 18 17:18:11 (at this point we dont really know if its -cpu related thou) Mar 18 17:19:18 sveinse: if you're using qemu for random emulation then yes, use qemux86-64 and turn on QEMU_USE_KVM Mar 18 17:19:38 maximal performance that way Mar 18 17:19:44 sveinse: as first test you can run the qemu manually and edit the cpu parameter Mar 18 17:21:08 rburton: thanks. Then I'll recompile for x64. It takes a while, so great to start in the evening Mar 18 17:33:32 rburton: are you saying i need both gtk+ and gtk+3 in my DEPENDS? Mar 18 17:33:36 no Mar 18 17:33:39 gtk+ is a recipe Mar 18 17:33:41 so is gtk+3 Mar 18 17:33:45 ok Mar 18 17:33:49 you depend on gtk+, which is gtk+ 2.something Mar 18 17:34:04 (which has been unmaintained for not a long way off a decade now) Mar 18 17:34:16 just change it to gtk+3 Mar 18 17:34:28 but when i changed the gtk+ to gtk+3 i am now getting "WARNING: wxwidgets-3.0.4-r0 do_package_qa: QA Issue: wxwidgets rdepends on gtk+, but it isn't a build dependency, missing gtk+ in DEPENDS or PACKAGECONFIG? [build-dep Mar 18 17:34:40 DEPENDS = "webkitgtk gstreamer gtk+3 jpeg tiff libpng zlib expat libxinerama libglu" Mar 18 17:34:48 so it still links to gtk2 Mar 18 17:34:53 you'll need to figure out how to change that Mar 18 17:35:35 is it perhaps this? BINCONFIG_GLOB = "arm-fslc-linux-gnueabi-gtk2-unicode-3.0" Mar 18 17:35:47 no Mar 18 17:35:53 i've forgotten what BINCONFIG_GLOB does Mar 18 17:36:00 thats the glob that tells the binconfig class where the scripts to rewrite are Mar 18 17:36:13 you'll want a lot more * in there Mar 18 17:36:30 here is the entire recipe now: https://paste.fedoraproject.org/paste/01VykQ~dpLwTgejHpEAoVA Mar 18 17:36:30 ie any build that isn't gtk2 for arm using freescale won't have that filename Mar 18 17:36:56 i'm guessing you need to look how to build wx and tell it what backend to use Mar 18 17:47:21 yates: if you haven't found it, AC_ARG_WITH(gtk, [[ --with-gtk[=VERSION] use GTK+, VERSION can be 4 (EXPERIMENTAL), 3, 2 (default), 1 or "any"]], Mar 18 17:54:14 rburton: i had not. are you saying this would be a separate line in the recipe file? Mar 18 17:54:46 yates: if you run ./configure --help for wxwidgets you'll see "--with-gtk=VERSION etc etc" Mar 18 17:54:53 so that tells you how to tell it to use gtk3 Mar 18 17:55:30 aha. Mar 18 17:56:18 i found that removing the webkitgtk DEPENDS caused the build to complete without that do_package_qa warning. do you think i still need to look into this? Mar 18 17:56:44 my conjecture was that webkitgtk itself DEPENDSed on gtk+ (2) Mar 18 17:57:01 i'll try it - what the heck Mar 18 18:04:31 RP: do you have a link with console output for the https://autobuilder.yocto.io/pub/non-release/20190318-7/testresults/ failure? Mar 18 18:12:36 Hi Does anyone know if there is a reason why the devtool build command doesnt run 'deploy' task? Mar 18 18:12:36 Right now its only running build task, populate sysroot, packagedata Mar 18 18:13:20 would there be any reasons to not stick deploy into the tasklist for devtool build? Mar 18 18:20:43 how do you force the rebuild of an indivisual recipe Mar 18 18:20:48 individual Mar 18 18:25:33 or just build an individual recipe Mar 18 18:29:43 black_13: bitbake Mar 18 18:30:29 can you provide the full path name Mar 18 18:30:38 no Mar 18 18:32:13 just the recipe name. if you're rebuilding svn, then "bitbake svn" Mar 18 18:32:39 the specified layers determine which specific recipe and bbappends (if any) are used Mar 18 18:40:32 yah Mar 18 18:42:27 black_13: to force a rebuild, just bitbake -c clean somerecipe; bitbake somerecipe. that won't necessarily rebuild from scratch, however, it'll pull from the sstate cache to use the prebuilt binaries. to rebuild from scratch, bitbake -c clean somerecipe; bitbake -C fetch somerecipe; is the most reliable method. there is -c cleansstate, but if sstate mirrors are involved, it may still download and use those rather thanj building from scratch, whereas -C Mar 18 18:42:27 will force the tasks to run Mar 18 19:22:54 RP: can I close https://bugzilla.yoctoproject.org/show_bug.cgi?id=12499 as "won't fix"? Mar 18 19:22:55 Bug 12499: normal, Low, Future, jon.mason, NEW , [2.5 M1 RC3] qemuarm: failed to shutdown Mar 18 20:30:25 rburton: modifying wx ./configure options... :) Mar 18 20:30:41 yates: no.... passing the right one via EXTRA_OECONF Mar 18 20:30:57 --help tells you the available options, pass the right one to configure Mar 18 20:31:05 rburton: does that work in morty? Mar 18 20:31:18 passing options to configure scripts has literally always worked Mar 18 20:31:24 ok, great Mar 18 20:31:52 thanks Mar 18 20:32:27 probably worth reading how autotools.bbclass works too Mar 18 20:34:21 For information on how this variable works within that class, see the meta/classes/autotools.bbclass file. Mar 18 20:35:23 wow, self-documenting code... Mar 18 20:36:17 Does yocto have any mechanisms for compiling an installable package using the SDK? Mar 18 20:38:40 Because it is nice and well to make recipes with packages in Yocto, yet Yocto and bitbake is not for everyone. Especially for those who are working with component development and are more or less ignorant to the workings of bitbake. The SDK provides a toolchain, and it does provide target deployment, but does it allow producing something that could be used as a package feed? Mar 18 20:39:59 sveinse: i like your idea Mar 18 20:43:13 New news from stackoverflow: How to add users at build time via Yocto? Mar 18 20:53:51 is there predifined file for the root directory for the bitbake recipe Mar 18 20:56:59 black_13: i don't understand your question. recipes are under /sources in various places Mar 18 20:57:21 sources/poky, sources/meta-mycompany, etc... Mar 18 20:58:39 there is a ${D} and ${system} is there something that represents /root Mar 18 21:14:18 black_13: you can find those types of definitions in <>/sources/poky/meta/conf/bitbake.conf Mar 18 21:14:48 not sure, to answer your question Mar 18 21:17:13 ${IMAGE_ROOTFS} is the answer Mar 18 21:20:39 JaMa: If you're still around: This is the config for Qt on MACHINE="emux86" https://bpaste.net/show/be5acd9f2fdb - Most noticable: "Building for: linux-oe-g++ (i386, CPU features: )" Mar 18 21:23:23 AVX .................................. AVX AVX2 Mar 18 21:23:42 you don't have AVX2 in emulated pentium2 I believe Mar 18 21:23:50 change the cpu parameter to see if it helps Mar 18 21:25:15 Thanks, I'll try. Mar 18 21:26:08 The Qt configure output is a little vague on what "Target compilers supports" really mean. Does it literally mean what the compiler supports, or does it imply what the output is build with... Mar 18 21:26:11 it says that target compiler supports these (not that it's actually going to use them, but lets assume it does which would explain the SIGILL, but doesn't explain why it doesn't show readable error about missing instructions like with ssse3 in my case Mar 18 21:27:09 it could be that in your case it crashes before it even reaches the point where it verifies the features of the cpu it's running on Mar 18 21:28:09 I'm attempting to make an image with qtbase-dbg Mar 18 21:29:40 changing -cpu is much faster than rebuilding anything, but your call Mar 18 21:31:14 psrcode: the console output is embedded in there, extracting it is a pain, I have a "todo" item for that :( Mar 18 21:31:39 psrcode: specifically its embeddedin https://autobuilder.yocto.io/pub/non-release/20190318-7/testresults/qemux86-64-ptest/testresults.json Mar 18 21:32:12 psrcode: I'll get it out soon, just several things to do right now (including sleep) :/ Mar 18 21:32:29 jonmason: yes, we can close that now Mar 18 21:35:01 JaMa: Yeah, that did work. I picked -cpu Haswell by random, and now the Qt app is running. Thanks Mar 18 21:35:31 Not sure how high up I need to go for this thou. My insight into the intel gens are limited Mar 18 21:38:03 But as rburton suggested if I understood him correctly, running a qemux86_64 when my host is running a 64-os is perhaps more efficient, so I'll attempt that. Mar 18 21:38:19 ${WORKDIR} represents what Mar 18 21:41:56 sveinse: yes, qemux86-64 can be more efficient but the default -cpu might be too high for your host (like it was for me in the e-mail thread I've shared before) Mar 18 21:47:13 why am i getting this package qa error? i don't remember getting it previously. Mar 18 21:49:04 JaMa: yep. got it. Mar 18 22:01:05 THIS error: https://paste.fedoraproject.org/paste/X3zRAOZ-34AzWwmOAVOVTQ Mar 18 22:13:58 because /usr/lib/wx/3.0/web-extensions/webkit2_extu-3.0.so doesn't belong to -dev package Mar 19 02:43:01 Where are the cross-compiler tools located to build the Yocto project? **** ENDING LOGGING AT Tue Mar 19 02:59:57 2019